ChatGPT has become a vital tool for millions, but new research suggests that depending too much on the AI chatbot might actually make users feel lonelier. Two recent studies—one by MIT’s Media Lab and another by OpenAI—explored how interactions with ChatGPT, particularly through its voice mode, impact users’ emotional well-being.
The MIT study involved 1,000 participants using ChatGPT over four weeks, while OpenAI analysed 40 million chatbot interactions. Across both, researchers found that users with strong emotional attachment tendencies were more likely to experience loneliness, while those who trusted ChatGPT more tended to develop emotional dependence.
The effect on power users
The research also pointed out that ChatGPT’s most frequent users—often referred to as “power users”—were more likely to consider the chatbot a friend or even assign it humanlike emotions. Additionally, those who engaged in personal conversations with the AI were found to experience higher levels of loneliness over time.
Interestingly, while voice-based chatbots initially seemed to reduce feelings of loneliness more effectively than text-based interactions, this benefit faded with excessive use, especially when the chatbot had a neutral-sounding voice.
AI as a substitute for human connection?
With 400 million weekly users, ChatGPT has increasingly become a go-to for personal advice and companionship, despite concerns from mental health professionals. A 2024 YouGov survey found that over half of young Americans (18-29 years old) felt comfortable discussing mental health concerns with AI.
Some even believe ChatGPT provides better personal advice than human columnists. But as AI-driven companionship services like Replika and Character.ai face growing scrutiny, researchers are calling for further investigation into how chatbot design—especially factors like voice expressiveness—affects user behaviour and emotional health.
Ultimately, the question remains: Is ChatGPT helping users cope with loneliness, or deepening it?
